{"slug": "white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday", "title": "White House Finalizes Secret AI Model Evaluation Framework, Schedules Tuesday Industry Briefing", "summary": "The White House confirmed it met its August 1 deadline to finalize a voluntary AI model evaluation framework under Executive Order 14409, but the framework's benchmarks and thresholds are classified and have not been released publicly. OpenAI, Google, and Anthropic reviewed a draft in late July, and a staff-level briefing with those companies plus Meta is scheduled for Tuesday. The framework establishes a 30-day pre-release government review window for frontier models with advanced cybersecurity capabilities, with the NSA playing a central role.", "body_md": "# White House Finalizes Secret AI Model Evaluation Framework, Schedules Tuesday Industry Briefing\n\n- The White House confirmed it met its August 1 deadline to finalize a voluntary AI model evaluation framework under Executive Order 14409, signed June 2\n[[1]](https://www.axios.com/2026/08/03/white-house-finalizes-ai-framework-behind-closed-doors) - The framework's benchmarks and model thresholds are classified, and the framework itself — while unclassified — has not been released publicly\n[[2]](https://thenextweb.com/news/white-house-ai-framework-secret-voluntary-classified) - OpenAI, Google, and Anthropic reviewed a draft in late July and submitted edits; a staff-level briefing with those companies plus Meta is scheduled for Tuesday\n[[3]](https://www.yahoo.com/news/politics/articles/white-house-finalizes-voluntary-ai-162700622.html) - The framework establishes a 30-day pre-release government review window for frontier models with advanced cybersecurity capabilities, with the NSA playing a central role in assessments\n[[4]](https://www.nortonrosefulbright.com/en/knowledge/publications/900af3cf/executive-order-establishes-voluntary-early-access-framework-to-frontier-ai-models) - Critics say the secrecy around the framework contradicts transparency promises and creates a de facto gating mechanism companies cannot publicly evaluate\n[[2]](https://thenextweb.com/news/white-house-ai-framework-secret-voluntary-classified)\n\nThe White House said Monday it completed a voluntary framework for evaluating advanced AI models by the August 1 deadline set in President Trump's June 2 executive order on AI security — but it will not disclose what the framework contains, who has reviewed it in full, or when companies will begin operating under it [1].\n\nThe framework, mandated by Executive Order 14409, establishes criteria for determining whether AI models under development qualify as \"covered frontier models\" subject to a 30-day government review window before release. The benchmarks used to assess models' cybersecurity capabilities and the thresholds that trigger review are classified [[2]](https://thenextweb.com/news/white-house-ai-framework-secret-voluntary-classified) [4].\n\nAnthropic, Google, Meta, and OpenAI are scheduled to attend a staff-level briefing at the Office of the National Cyber Director on Tuesday to review the completed framework. Google, OpenAI, and Anthropic jointly reviewed a draft and submitted edits in late July, though it is unclear how many other companies were consulted [3].\n\n\"The voluntary framework outlined in the June 2nd executive order was complete by the deadline. Discussions with industry about next steps are underway,\" a White House official said [3].\n\n## What the Framework Does\n\nExecutive Order 14409, titled \"Promoting Advanced Artificial Intelligence Innovation and Security,\" directed federal agencies to create a voluntary mechanism for frontier AI developers to provide the government early access to models before release. Under the framework, the federal government receives access to covered models for up to 30 days before they are released to other trusted partners. During that window, AI developers and government agencies collaborate on selecting which external partners gain early access [4].\n\nThe order explicitly states it \"does not create any sort of mandatory licensing, pre-clearance or permitting mechanism\" [4]. The NSA plays a central role in assessing models' advanced cyber capabilities, while the framework pairs with Gold Eagle, a public-private AI cybersecurity clearinghouse launched July 14 that coordinates vulnerability discovery and remediation across critical infrastructure\n\n.\n\n[[5]](https://www.whitehouse.gov/releases/2026/07/white-house-launches-gold-eagle-initiative-for-unprecedented-cybersecurity-vulnerability-coordination/)The EO directed agencies to develop a classified benchmarking process — involving the NSA, CISA, and NIST — to assess covered models, but does not publicly specify the technical criteria that would designate a model as \"covered\" [4]. As Norton Rose Fulbright noted in its analysis: \"developers may have limited public guidance, at least initially, on how the government will assess a model's cyber capabilities\"\n\n.\n\n[[4]](https://www.nortonrosefulbright.com/en/knowledge/publications/900af3cf/executive-order-establishes-voluntary-early-access-framework-to-frontier-ai-models)## Why the Secrecy Matters\n\nThe framework's opacity has drawn criticism from AI safety advocates and policymakers who have been waiting to see what rules apply to the most powerful models in the world. A White House official defended the approach: \"Just because things are unclassified that doesn't mean we are going to broadcast them to everyone\" [2].\n\nCritics argue that building a framework on undisclosed benchmarks and secret thresholds contradicts the executive order's promised transparency. The combination of classified components and government preview authority creates what observers have called a \"de facto gating mechanism\" that companies cannot publicly evaluate or challenge — despite the framework's nominally voluntary status [2].\n\nThe lack of publicly defined thresholds has created regulatory uncertainty for frontier AI labs. Companies including OpenAI, Anthropic, Google, Microsoft, and xAI face difficulty determining whether their model architectures trigger \"covered frontier model\" designations, which some industry participants say forces delayed release timelines [6].\n\n## Industry and Geopolitical Context\n\nThe framework arrives amid an accelerating global AI race. While the U.S. works to establish its oversight regime behind closed doors, China's DeepSeek is actively building 1GW data centers in Mongolia without comparable regulatory constraints, a dynamic that has heightened competitive concerns within the administration [6].\n\nThe executive order itself was prompted by the K3 Cyber incident and intersects with separate legislative efforts including the Kill Switch Act. Anthropic CEO Dario Amodei has publicly advocated for frameworks governing frontier models, and his company is among those most closely engaged with the White House process [6].\n\nThe voluntary nature of the framework reflects the administration's stated preference for a light-touch approach to AI regulation, but the classified benchmarks and 30-day government review window give the program more teeth than a purely advisory role. Policymakers and U.S. allies have been closely watching the framework's development as a potential model for international AI governance [1].\n\n## What's Next\n\nTuesday's staff-level meeting at the Office of the National Cyber Director will be the first formal industry briefing on the completed framework. It remains unclear whether the framework will eventually be made public, or whether additional companies beyond the four attending Tuesday's meeting will be brought into the process [3].\n\nThe administration must also deliver on two other mandated deliverables from the executive order that were due by August 1: a classified benchmarking process involving the NSA, CISA, and NIST, and a federal cyber workforce expansion plan from the Office of Personnel Management. As of the deadline, it was unclear whether those had been completed [6].\n\n## Companies mentioned\n\n## Further sources\n\n[[1] Axios: White House finalizes AI framework behind closed doors ↗](https://www.axios.com/2026/08/03/white-house-finalizes-ai-framework-behind-closed-doors)\n\n[[2] The Next Web: The White House says its AI framework is done. It will not say wh… ↗](https://thenextweb.com/news/white-house-ai-framework-secret-voluntary-classified)\n\n[[3] Yahoo News: White House finalizes voluntary AI oversight framework ↗](https://www.yahoo.com/news/politics/articles/white-house-finalizes-voluntary-ai-162700622.html)\n\n[[4] Norton Rose Fulbright: EO sets voluntary 'early access' framework for AI models ↗](https://www.nortonrosefulbright.com/en/knowledge/publications/900af3cf/executive-order-establishes-voluntary-early-access-framework-to-frontier-ai-models)\n\n[[5] White House: Gold Eagle Initiative for Unprecedented Cybersecurity Vulnerabilit… ↗](https://www.whitehouse.gov/releases/2026/07/white-house-launches-gold-eagle-initiative-for-unprecedented-cybersecurity-vulnerability-coordination/)\n\n[[6] Yahoo Finance: White House AI Framework Deadline Lapses Without Public Delivera… ↗](https://finance.yahoo.com/technology/ai/articles/white-house-ai-framework-deadline-002011007.html)\n\nThe stories that matter, in one email. Free — unsubscribe anytime.", "url": "https://wpnews.pro/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday", "canonical_source": "https://mlq.ai/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday-industry-briefing/", "published_at": "2026-08-04 12:23:04.167179+00:00", "updated_at": "2026-08-04 12:23:06.586691+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety"], "entities": ["White House", "OpenAI", "Google", "Anthropic", "Meta", "National Security Agency", "Office of the National Cyber Director", "Executive Order 14409"], "alternates": {"html": "https://wpnews.pro/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday", "markdown": "https://wpnews.pro/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday.md", "text": "https://wpnews.pro/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday.txt", "jsonld": "https://wpnews.pro/news/white-house-finalizes-secret-ai-model-evaluation-framework-schedules-tuesday.jsonld"}}